MD2 vs. MD3 vs. HL-LIFEModels vs. UT-Models...

Hi again,
does anyone use this gameformats for his models in his 3D-scenery ?
Can anyone explain me to load MD3files - i think this is the best fileformat for models (object`?) in a 3D-world - but i´m not able to find any description !!!
Does anyone know any solution ??
Or is the MD2 format “enough”, i mean technically featured ??? Sometime ago, i read that the maximum number of polygons in the MD3 format is the same as in the MD2 format specification - BUT: the modellers use more polygons, caus they believe they can use more - is this true ??
Or should i use HalfLife-Models, or UnrealTournament-Models ?!?!?!?!?!?
Please help…

DJS

first of all, neither md2 or md3 files are halflife models…md3 is quake 3. second of all, either way you look at it, quake 3 is a hell lot better that ut, so i’d go with md3.

Umm…

I’d say that .max or .ma files would be the
ideal container formats if you want richness
of the model (3DSMax and Maya native files,
respectively).

One rung down is .3ds; still fairly rich and
with a number of different shading parameters
and animation methods for your model. You can
add special chunks for skeletal animation,
or just use some specially named material.

.md2 and .md3 are tuned specifically for the
engines they were developed for (Quake 2 and
3, respectively) although I hear .md3 has
slightly more extensible support than .md2
and could support skeletal animation even if
Quake 3 doesn’t use it. Of course, something
that’s not used is not tested, so it’s not
REALLY implemented…

Looking at my .md2 file parser, it can deal
with pretty much any number of triangles per
mesh, although the more serious limitations
are the 65536 maximum vertexes, and the
quantization of vertex coordinates to 8 bits
of resolution, and the even coarser
quantization of normals from a predetermined
look-up table. Oh, and the texture s/t
coordinates seem quantized to whole pixel
offsets, too, which might be construed as a
limitation.

Now, if you’re worried about expressive
power, design your own format and write an
exporter or converter from the format from
wherever you’ll be getting your models. Or
just load your modeller’s output directly.
Or, worst case, just do a web search for
.md2 and .md3 file formats versus .3ds or
others, and judge for yourself.

Yes, im thinking of writing my own format - but i have no idea, how to do the texture-managing for each surface so easy, so that the user can map it with comfort ! Ok, i must say, i have never used a Q2/3 Editting-tool, like QRadiant, for skinning - i know nothing about these tools and how they works - BUT: it cannot be possible that you must skin each single face with a part of your texture by selecting the needed surface !!! this must be a lie !! so i need a way to calculate the texture on objects or on objectparts. I have a model on my harddisk that is a airplane (military) - the complete texture for the whole plane is saved in one file. Worked the Designer as followed subsribed(?): he build the model, takes a lot of textures, select for each face, the texture he want - the program writes a file which puts all texture-parts together !!! is this the way ???
This cannot be !!!
There must be a way to calculate the texture automatically !!! or not ??

DJS

How the texture coordinates get into the
model files is a tools issue, not a display
issue. There are several stand-alone tools
and plug-ins which intend to make this
process easier; either by letting you “paint”
on the model directly, or by using some
heuristic to try and spread the triangles of
the mesh somewhat evenly over the texture.
3DSMax only has the simplest coordinate
mechanisms available internally: spheric,
cylindric, box, plane etc. Anything fancier
requires something extra.

The problem is that with many “automatic”
approaches is that they may either get fairly
low texture utilization (lots of unused parts
of the texture file) or they will stretch
some triangles so that, say, a very wide but
short strip of pixels map to a “normal-
shaped” triangle, leading to display
artifacts. There’s no substituting a real
human brain (yet).

What, you thought 3D modeling was easy?
There’s a reason good modelers are in demand
you know :slight_smile:

You said something about “direct model paint” can i imagine this as, like “face skinning”: I HAVE TO SELECT EVER FACE. NEXT STEP IS TO SELECT THE ARE OUT OF THE TEXTURE, YOU WANT TO MAP. NEXT STEP IS, TO ROTATE AND STRETCH THE SELECTED TEXTURE AREA TO A FITTING SIZRE ??? AAARRRRRRRRRGH!!!

IF I BUILD UP MY TOOL FOR CREATING 3DENVIRONMENTS WITH THIS SKINNING “WAY” - EVERY USER WOULD KILL ME !!! or not ?

mmmh…then a modeler must need for every model more than 2 weeks !!!
or am i wrong ?

DJS

OKOK,
as you have read in the other discussion, i´m running about the 3DS-loader-problem on the message board, both discussion are going out on the same problem.
I think i should do symbiosis of both:
For terrain and large objects, i going to use 3DS-data, this data must be skinned in my tool called “MEDIACreator” by hand. For smaller objects, animations and humans im going to use MD2/MD3, caus texture-coordinate data is allready included: so the user hasn´t to play around with small textures on tiny surfaces, which are so small, that he can´t see on the screen… :slight_smile:
My tool will put this data together in an own, engine internal format. which is optimized for my engine…

I think this is the best way !

DJS

[This message has been edited by DJSnow (edited 12-14-2000).]

Wait…
you said something about your MD2 parser - do you have a MD3 parser, too ?
or some information about the MD3 format ?
i´ve hvae found nothing !

DJS

if ure having probs with the md2 dont even think about the md3 format.
about the making 2d texture that maps onto the 3d mesh i agree AAAAARRRRRGGG its the most time consuming and boring thing. its someit i dont understand why cant 3dmax or someit else do this 4 u. of course it wont be perfect but it will be 90% good now its like make a mopdel in 30min + then spend 3 hours getting the mapping coords from it !!!

3DSMax (and other tools) do do some
texture coordinate setup for you. It’s the
UVWMap modifier, right there on the modifier
page. The problem is, it’s not very good.
There may be plug-ins which do it better, but
you’ll probably still need hand-tuning in
some program which allows you to re-map
texture coordinates.

A good modeler can do a typical humanoid in
about three days. (that’s three days of
eight hour work) So “two weeks” sounds
excessive – if you know what you’re doing.

3ds max also has a modifier called unwarp uvw, one of the best features of this app for low poly modelers, and u should try getting deep paint 3d.

If using 3ds max, there are two invaluable tools for skinning meshes. The first one is Texporter, which is generates mesh outlines for you with gradient colors showing the topology of the mesh. The seond one is Chilliskinner, which is a tool for breaking apart, arranging and flattening the the mesh, and later recombine it. With those two tools, skinning is considerably easier/faster.

Bringing back the dead huh? I guess we can from now on call you Jesus. :smiley:

-SirKnight

Ehr, yeah, kindof forgot that it was an old thread i found by searching… Tabbed browsing requires a strong mind appeareantly.

Please ignore the “quakeish” use of the term “skinning”. Those tools however do are really useful, I would never leave home without em.

Btw, “Jesus” was a very polite term. “Foul necromancer” is another candidate :smiley: .

chilliskinner would be an outdated technique, since max5 the unwrap tools got so mighty that the “old way of breaking the mesh” isnt really called for anymore

the statement about 3 days for a human, also doesnt imply what kind of detail your going after…

about 3d formats, the most important point is what it’s for, is it for level geometry, actors, can it contain animation, or should be an “all in one” type of scene representation.
coming from the artists view, md2 and md3 would not be good picks anymore, they are just too limiting or rather complicated to export properly, especially md2 is severely outdated. HL’s modelformat is much more mighty when it comes to animation as it is skeletal. But in general I would look around more what other possibilties there are and sometimes it’s wise to do your own format, cause then you get specifically what you need.

well it’s really mostly a “what you need for, and what can your engine do with it” question.

Greetings! I have asked the same question over and over again, so let’s go:

MD2 - Quake 2, Only ONE PCX per Mesh, very easy to handle load, FIXED animation, nothing more to do

MD3 - Quake 3, Multi texture, 3 Layers, can rotate Head, upper & lower body in seperate directions, NO BONES, easy to handle

MDL/SMD - Half-Life, Very easy to load, Skeletal animation, multi texture & multi object support, can define in many ways. Easy to export. The exporter compiles only 8BIT 256 color textures but stores them as 24 BIT RGB bmps into MDL files.

For me, I use Unreal 2003 PSK and PSA files. Easy to load, easy to render, Use skeletal animation and they are easy to export from 3DSMAX 5 or 6 and no problems at all.

Originally posted by bgl:
I’d say that .max or .ma files would be the
ideal container formats if you want richness
of the model (3DSMax and Maya native files,
respectively)
The maya ascii file is simply a set of mel commands that recreate all the nodes and connections within the scene. Short of writing maya from scratch you would not be able to get much use out of it.

A better alternative for Maya would be something like this :

Maya Exporter Factfile

Originally posted by Xarragon:

Btw, “Jesus” was a very polite term. “Foul necromancer” is another candidate :smiley: .

I’ll have to remember that. :smiley:

-SirKnight

Just a side note, md3 DOES use bone animation, it just happens that its single weight, and thus the joints are static rather than deforming as they are with weights and multiple bones (skinning), but while were on the topic, anyone mind telling me which way a single dimension array would be in order for openGL to interpret it the way i want?

i thought it was

0 1 2 3
4 5 6 7
8 9 10 11
12 13 14 15

but i saw something that seemed to conflict with that the other day, and since i have yet to test out my matrix class other than giving results and checking them against my calculator

and finnaly, in openGL where is the transformation vector?

the site or the bottom?

thanks
-Dan

Originally posted by Ademan:
[b]and finnaly, in openGL where is the transformation vector?

the site or the bottom?
[/b]
bottom, 12,13,14